One of the main issues confronting the semiconductor processing industry is that of the capacitive-resistance problem in metallization layers. An industry-wide effort has undertaken to address the problem. Since the beginning, the semiconductor processing industry has relied on aluminum and aluminum alloys to serve as metallization layers. Silicon dioxide was selected as the insulator of choice although polyimide, a polymer, was used in a number of products by IBM for a number of years. With each succeeding generation of technology, the capacitive-resistance problem grows. Because each generation requires that the dimensions of the semiconductor structure be reduced, the minimum line-space combination must also decrease. As the line-space combination decreases, the capacitance and resistance of the semiconductor structure increases. Thus, these increases contribute to the problem.
Copper metallurgy has been proposed as a substitute for aluminum metallurgy as a material for the metallization layers since copper exhibits greater conductivity than aluminum. Yet several problems have been encountered in the development of copper metallurgy. The main issue is the fast diffusion of copper through an insulator, such as silicon dioxide, to form an undesired copper oxide compound. Another issue is the known junction-poisoning effect of copper. These issues have led to the development of a liner to separate the copper metallization layer from the insulator. The use of titanium nitride as a liner was proposed by C. Marcadal et al., “OMCVD Copper Process for Dual Damascene Metallization,” VMIC Conference Proceedings, p. 93-7 (1997). The use of tantalum nitride as a liner was proposed by Peijun Ding et al., “Copper Barrier, Seed Layer and Planarization Technologies,” VMIC Conference Proceedings, p. 87-92 (1997). The use of titanium as a liner was proposed by F. Braud et al., “Ultra Thin Diffusion Barriers for Cu Interconnections at the Gigabit Generation and Beyond,” VMIC Conference Proceedings, p. 174-9 (1996). The use of tungsten silicon nitride as a liner was proposed by T. Iijima et al., “Microstructure and Electrical Properties of Amorphous W—Si—N Barrier Layer for Cu Interconnections,” VMIC Conference Proceedings, p. 168-73 (1996). The use of zirconium, hafnium, or titanium as a liner was proposed by Anonymous, “Improved Metallurgy for Wiring Very Large Scale Integrated Circuits,” International Technology Disclosures, v. 4 no. 9, (Sep. 25, 1996). The use of titanium as a liner was proposed by T. Laursen, “Encapsulation of Copper by Nitridation of Cu—Ti Alloy/Bilayer Structures,” International Conference on Metallurgical Coatings and Thin Films in San Diego, Calif., paper H1.03 p. 309 (1997). The use of tantalum, tungsten, tantalum nitride, or trisilicon tetranitride as a liner is currently favored by the industry. See Changsup Ryu et al., “Barriers for Copper Interconnections,” Solid State Technology, p. 53-5 (1999).
Yet another solution to the problem of fast diffusion of copper through an insulator was proposed by researchers at Rensselaer Polytechnic Institute (hereinafter, RPI). See S. P. Muraka et al., “Copper Interconnection Schemes: Elimination of the Need of Diffusion Barrier/Adhesion Promoter by the Use of Corrosion Resistant, Low Resistivity Doped Copper,” SPIE, v. 2335, p. 80-90 (1994) (hereinafter, Muraka); see also Tarek Suwwan de Felipe et al., “Electrical Stability and Microstructural Evolution in Thin Films of High Conductivity Copper Alloys,” Proceedings of the 1999 International Interconnect Technology Conference, p. 293-5 (1999). These researchers proposed to alloy copper with a secondary element, which is either aluminum or magnesium. In their experiments, they used copper alloys with at least 0.5 atomic percent aluminum or 2 atomic percent magnesium. When the copper alloy is brought near the insulator, silicon dioxide, the secondary element and silicon dioxide form dialuminum trioxide or magnesium oxide. The formed dialuminum trioxide or magnesium oxide acts as a barrier to the fast diffusion of copper into the silicon dioxide.
Along the same technique as proposed by RPI, Harper et al. discuss in U.S. Pat. No. 5,130,274 (hereinafter, IBM) the use of a copper alloy containing either aluminum or chromium as the secondary element. As above, the secondary element with the insulator, such as silicon dioxide or polyimide, forms a barrier to the fast diffusion of copper.
Semiconductor products with some of the discussed solutions to the fast diffusion of copper have begun to ship, on a limited basis, and yet the problem of reducing the resistivity in ever smaller line dimensions is still present. It has been shown by Panos C. Andricacos, “Copper On-Chip Interconnections,” The Electrochemical Society Interface, pg. 32-7 (Spring 1999) (hereinafter Andricacos), that the effective resistivity obtainable by the use of barrier layers was approximately 2 microhm-centimeters with a line width greater than 0.3 micrometer. The effective resistivity undesirably increases for lines narrower than that. The alloy approach investigated by RPI had similar resistivity values as found by Andricacos. RPI also found that the use of 0.5 atomic percent aluminum, in the copper, was apparently insufficient to give complete protection from copper diffusion into the silicon dioxide although a significant reduction in the rate of copper penetration through the silicon dioxide was achieved. It should be noted that the maximum solubility of aluminum in copper is 9.2 weight percent or approximately 18 atomic percent whereas the maximum solubility of magnesium in copper is 0.61 weight percent or approximately 0.3 atomic percent. Thus, the alloys used by RPI were saturated with magnesium but far below the saturation limit when aluminum was used as the secondary element in the alloy.
Other researchers have focused on the capacitive effect. The capacitive effect has been studied with respect to polymers, such as polyimide, which are used to substitute for silicon dioxide as insulation in semiconductor structures. Some of these polymers have dielectric constants that are considerably lower than silicon dioxide, and a presumption can be made that the use of these polymers should lessen the undesired capacitive effect. Yet, when one of these polymers is cured to form an insulator near the vicinity of the copper metallization layer, the polymer reacts with the copper metallization layer to form copper dioxide, a conductive material. See D. J. Godbey et al., “Copper Diffusion in Organic Polymer Resists and Inter-Level Dielectrics,” Thin Solid Films, v. 308-9, p. 470-4 (1970) (hereinafter, Godbey). This conductive material is dispersed within the polymer thereby effectively raising the dielectric constant of the polymer and in many cases even increasing its conductivity. Hence, the undesired capacitive effect continues even with the use of lower dielectric polymer materials.
Andricacos points out that the use of copper along with cladding offers a significant improvement in conductivity over the titanium/aluminum-copper alloy/titanium sandwich structure now in widespread use throughout the industry. Andricacos also noted that as the line width decreases even a thin liner would undesirably effect the line resistance. The proposals by RPI and IBM attempt to address this problem by forming the liner using a copper alloy. The liner formed using a copper alloy displaces a portion of an area that was occupied by the insulator.
However, in solving one problem, RPI and IBM introduce another problem. The copper alloys used by RPI and IBM essentially lack the desirable properties of copper that initially drove the industry to use it. As was pointed out by RPI, the use of an alloy containing aluminum, even at a concentration so low as to not be completely effective in preventing the diffusion of copper, shows a measurable increase in resistance. IBM used only one layer of the alloy. Yet, that one layer has a high concentration of aluminum and will undoubtedly have an undesired effect on the resistivity.
As the minimum dimensions shrink, the use of even a twenty-Angstrom layer of an alloy with higher resistivity will have a significant effect on the total resistivity of the conductor composite. For example, a 200-Angstrom film on both sides of a 0.1 micron trench is 40 percent of the total trench width. Therefore, at the same time that the dimensions of the metallization layer decrease, the specific resistivity undesirably increases.
It has also been shown that there is a significant difference between the amount of the undesired copper oxide compound that is formed when a polyimide insulator is used if the acidity of the polymer solution is low. This is the case if the precursor used in the formation of the polyimide is an ester instead of acid. In the case of PI-2701, which is a photosensitive polyimide that starts from an ester precursor, the amount of oxide formed is reduced by a factor of approximately four as compared to films with a similar final chemistry. See Godbey. It is thought that the slight acidity of PI-2701 may come from the photo-pac or the process used to form it. The films in the study by Godbey were all prepared by curing the liquid precursor in air or in an approximately inert environment. It is also well known that copper oxide will not form in and can be reduced by a high purity hydrogen atmosphere.
Muraka opines that the use of titanium as a barrier layer was found to increase the resistivity of the copper film significantly when heat-treated at temperatures of 350 degrees Celsius or above. If the heat-treatment was carried out in hydrogen, no increase in resistivity was reported. As this temperature is above the eutectoid temperature of the titanium-hydrogen system, the formation of titanium hydride is assumed to have occurred. Muraka also asserts that a similar increase in resistivity is seen with zirconium and hafnium containing copper alloys, yet Muraka provides no data to support the assertion.
Other research results weaken the conclusion of Muraka. See Saarivirta 1; see also U.S. Pat. No. 2,842,438 to Matti J. Saarivirta and Alfred E. Beck (Jul. 8, 1958). If one looks at the equilibrium phase diagrams of the copper-titanium and copper-zirconium systems, it can be seen that the solubility of zirconium in copper is more than ten times less than that of titanium. See Metals Handbook, v. 8, p. 300-2 (8th Ed.). It should also be noted that a series of copper-zirconium alloys have been disclosed that have quite good electrical conductivity.
It has been shown that alloys containing more than about 0.01 weight percent zirconium have a significant loss of conductivity in the as-cast state. See Matti J. Saarivirta, “High Conductivity Copper-Rich Cu—Zr Alloys,” Trans. of The Metallurgical Soc. of AIME, v. 218, p. 431-7 (1960) (hereinafter, Saarivirta 1). It has also been shown that the conductivity of even a 0.23 weight percent zirconium alloy is restored to above 90 percent of IACS when the alloy, in the cold drawn state, is heat-treated above 500 degrees Celsius for one hour. This shows that a significant amount of the zirconium, which was in solid solution in the as-cast state, has precipitated as pentacopper zirconium. From this data, it can be seen that if the zirconium content in the copper is kept low the conductivity of the resulting metallurgy can be above 95 percent of IACS. If it is desired to deposit a zirconium layer on top of a copper layer the temperature of deposition of the zirconium should be kept below 450 degrees Celsius, such as between 250 degrees Celsius and 350 degrees Celsius. Such deposition may occur in a single damascene process or at the bottom of vias in a dual-damascene process. The term “vias” means the inclusion of contact holes and contact plugs. When the deposition temperature is kept in this range, a thin layer of pentacopper zirconium tends to form initially thus inhibiting the diffusion of zirconium into the copper. While even at 450 degrees Celsius the solubility is low enough to give very good conductivity, and although zirconium and titanium have many properties that are very similar, their solubility in copper differs by more than a factor of ten. Therefore, the use of zirconium is much preferred over titanium for this application.
What has been shown is the need of the semiconductor processing industry to address the issue of interconnecting devices in integrated circuits as these circuits get smaller with each generation. Although aluminum was initially used as the metal for interconnecting, copper has emerged as a metal of choice. However, because of the fast diffusion of copper into the semiconductor insulator, the capacitive-resistive problem becomes an important issue that must be addressed. One solution is to use a liner, but with the reduction in the geometry of the circuits, the dimensions of the liner become inadequate to prevent the fast diffusion of copper. Another solution is to form a barrier material from the insulator and a copper alloy; this solution seems promising at first, but because the copper is alloyed, the desirable conductivity property of copper is diminished.
Thus, what is needed are structures and methods to inhibit the fast diffusion of copper so as to enhance the copper metallization layer in a semiconductor structure.